AUTHOR Dr Frank Reichert, Leibniz Institute for Educational Trajectories, Bamberg (Germany) TITLE Political Knowledge and Political Action: The Mediating Effect of Internal Political Efficacy 1 Work in progress please do not quote or cite without author s written permission! Comments welcome! ABSTRACT Knowledge about politics is often considered a precondition for political action. However, from an action theoretic point of view it is maybe not political knowledge but internal political efficacy that initiates political action. This paper hypothesizes that political knowledge is an important predictor of legal political activity, but fully mediated by internal political efficacy. This hypothesis is examined using a sample of university students and, in addition, data from the German Longitudinal Election Study. Initial results indicate that political knowledge is of some importance with regard to voting, although it does not have a direct effect on conventional political action when controlling for sociodemographics and internal political efficacy which appears as a mediator. These results are discussed considering limitations of both studies, the difference between habitual and reflective actions, and the importance of political education. KEYWORDS Education, Knowledge, Political Participation, Political Psychology, Political Sociology, Voting 1 Paper prepared for oral presentation at the Graduate Student Conference of the European Consortium for Political Research (ECPR), 3 5 July 2014, Innsbruck, Austria (scheduled time: Friday, 4 July 2014, 9:00 10:30). This research was supported by research grants from the Deutsche Forschungsgemeinschaft (DFG) to Bernd Simon, Christian-Albrechts-Universität Kiel (SI 428/13-3, -4). The author is also grateful to Sarah Trehern for her assistance in data collection.
Political Competences and Political Action: On the Mediation of Political Knowledge 2 INTRODUCTION Political knowledge has gained broad attention in the United States as a topic of research in political science, whereas in Germany, interest in political knowledge is still emerging (e.g., Maier 2009; Westle 2011). What seems clear is that the knowledge of politics among most people is insufficient to meet the standards of a competent citizen (e.g., Delli Carpini and Keeter 1996; Maier 2000; Westle 2013), even though political knowledge may be considered a significant quality of politically active and involved citizens. Since many citizens are politically active, a question is whether we could measure internal political efficacy as a proxy of political knowledge in survey research, i.e. if rather a subjective assessment of one s own political competences might be a mediating variable with respect to political knowledge. A study that examines this question was conducted earlier (cf. Reichert 2013). It will be summarized in brief and is extended using the long-term panel from the German Longitudinal Election Study (GLES; Rattinger et al. 2012). In addition to the parameters of the first study, political interest will also be taken into account in supplementary analyses of the replication study. THEORETICAL CONSIDERATIONS Kinds of Political Participation and Their Predictors Political participation consists of every activity a citizen takes voluntarily to influence authoritative or generally binding regulations and decisions on any of the different levels of the political system (Kaase 1992). The term behavioral politicization comprises actual political participation as well as the intention to act in a certain political way. Behavioral intentions are strong predictors of actual behavior (Ajzen 2001) and thus are worthy of analysis, too. The correlational relationship between intention and action is, however, by no means perfect (e.g., Ajzen 1991; Fishbein and Ajzen 1981; Gaiser and Rijke 2010; Steinbrecher 2009). The analyses presented in this article distinguish four kinds of political participation (c.f. Barnes et al. 1979; Gabriel and Völkl 2008; Steinbrecher 2009): Electoral political participation (voting) does not require intense effort, nor is it bound to a strong commitment. The only constraint on voting is formal regulations (e.g., citizenship). Conventional or traditional political activities are often institutionalized and are sometimes called party politics (e.g., supporting an election campaign). Unconventional activities refer to a broad range of less time-intensive or committed political participation (e.g., distributing leaflets). Finally, there exist non-normative, illegal political activities, which are located outside the legal framework (e.g., attending a violent demonstration).
Political Competences and Political Action: On the Mediation of Political Knowledge 3 In the following, only variables at the individual level will be considered to explain politicization; the societal or institutional context will not be addressed. At the individual level, to which political knowledge, competences, and political interest belong, are three important approaches explaining political behavior (cf. Kaase 1992; Steinbrecher 2009; Verba, Schlozman, and Brady 1995): The standard socio-economic status model of political participation (e.g., Verba and Nie 1972; Brady, Verba, and Schlozman 1995) is applicable to all kinds of legally accepted political activities (Kaase 1990). According to this resource model, people with higher education and people with higher incomes or better jobs tend to be more active. The existence of a migration background is negatively associated with diverse forms of political participation, but differences in education can often explain this finding (e.g., Gaiser and Rijke 2010; Heß-Meining 2000). Furthermore, women are less likely to participate in politics, particularly with regard to voting, conventional, or illegal political action (e.g., Gabriel 2013; Gaiser and Rijke 2010; Kaase 1990). Likewise, the effect of age on political behavior depends on the kind of action. The likelihood of participating in other kinds of political behavior especially declines when people get older (e.g., Krimmel 2000; Watts 2001; see also, Verba and Nie 1972). The social capital approach assumes that political participation depends on variables like interpersonal trust, social contacts, and volunteering (cf. Putnam 1993; 2000). The kind of network is of course an important explanatory facet, and, in sum, the correlations between social capital and politicization are only of moderate size (Gabriel et al. 2002; Gabriel 2004; van Deth 2002; Verba, Schlozman, and Brady 1995). Motivational variables constitute a third approach. They are often also used to operationalize constructs of rational approaches, which assume that individuals become politically active if they think it is likely to satisfy their own needs to the greatest possible extent (Opp 2009; Steinbrecher 2009). These variables comprise individual values and attitudes, interest in politics, internal and external political efficacy, and political trust. Political interest and internal political efficacy especially constitute powerful and positive predictors of political action (e.g., Delli Carpini and Keeter 1996; Finkel 1985; Gallego and Oberski 2012; Galston 2001; Krampen 2000b; Krimmel 2000; van Deth 2001; Verba, Schlozman and Brady 1995; see also below). At the individual level, biological variables like, for instance, personality traits (e.g., Mondak et al. 2010; Quintelier 2012) or genetics (e.g., Fowler, Baker, and Dawes 2008; Hatemi et al. 2007) have also been taken into consideration recently. The Role of Cognitive Politicization Variables in Behavioral Politicization Political competence can be defined as the ability to understand, judge, and successfully influence politics and political facts (e.g., Gabriel 2008). Political knowledge is a precondition for acquiring other profound political competences (i.e.,
Political Competences and Political Action: On the Mediation of Political Knowledge 4 political analyzing and reasoning, capacity to act politically, and methodical abilities; cf. GPJE 2004; Krammer 2008; Richter 2008; Sander 2008). Thus, the possession and recollection of political knowledge must be regarded as a content-related competence consequently, political knowledge is a political competence itself (Weißeno 2009; compare also Hoskins et al. 2008; Richter 2008; Rychen 2004). Political competences can, however, be real or imagined. If imagined, they rely on a person s subjective assessment of himself or herself. Both political knowledge and political competence are important in explaining political participation. Objective political knowledge Objective political knowledge is the range of factual information about politics that is stored in long-term memory (Delli Carpini and Keeter 1996, p. 10). Although a classification of content dimensions is discussed by some authors (e.g., Delli Carpini and Keeter 1996; Jennings 1996; Johann 2012), no such differentiation is considered in the following. Galston (2001, p. 224) summarizes the state of the art quite well in writing that [c]ivic knowledge promotes political participation : More extensive knowledge about polity, politics, and policy are presumed to enable and encourage people to participate in political action. Delli Carpini and Keeter (1996) especially argue that the knowledge of politics influences other political competences and actual political participation. Several studies underline the positive role of political knowledge as a predictor for at least voting and conventional political participation (e.g., Amadeo et al. 2002; Hoskins et al. 2008; Krampen 2000b; Torney-Purta et al. 2001; Schulz et al. 2010), and Westle (2011; 2012) even finds its effects on unconventional political participation. When going beyond the personal resource of knowledge and taking into account action theoretical models of political participation, questions about mediators of objective political knowledge emerge: Could it rather be sufficient to measure the subjective component of political competence, or do political knowledge and internal political efficacy both have separate explanatory value when controlling for each other? Following Ajzen (1991), subjective behavioral control can be seen as equivalent to internal political efficacy. It might then be true that, controlling for each other, only internal political efficacy is a statistically significant predictor of political action, mediating the effect of political knowledge 2 : Maybe subjective control, convictions, and feelings of competence are stronger predictors than objective knowledge (cf. Ajzen 1991; 2001; Krampen 1991; 2000a; 2000b; Ryan and Deci 2000; Vetter and Maier 2005)? 2 Westle (2011), however, argues that political knowledge has, in fact, separate effects on political action when controlling for political interest and subjective political competence. Her results, however, rely on cross-sectional data, whereas the current study uses panel data.
Political Competences and Political Action: On the Mediation of Political Knowledge 5 Internal political efficacy The concept of self-efficacy is based on the distinction of outcome expectations ( a person s estimate that a given behavior will lead to certain outcomes ) and efficacy expectations ( the conviction that one can successfully execute the behavior required to produce the outcomes ) (Bandura 1977, p. 193). The second one is the estimation that the person himself or herself is able to act successfully to gain the objective aimed. With regard to politics, this concept comprises the feeling that one is capable of understanding political facts and processes and is able to influence politics successfully oneself (cf. Almond and Verba 1965; Balch 1974; Campbell, Gurin, and Miller 1954). Although often used interchangeable with the term self-concept of one s own political competence (e.g., Campbell et al. 1960; Koch 1993; Krampen 2000b; Vetter and Maier 2005), the common term is internal political efficacy. Krampen (2000b, p. 290) states that the variables of frequency of political activity in everyday life, selfconcept of political competence, and political knowledge in adolescence are the most significant discriminators for voting versus nonvoting behavior of young adults. Numerous studies show that subjective political efficacy plays a significant, positive role in behavioral politicization (e.g., Krampen 1991; 2000b; Preiser and Krause 2003; Schulz et al. 2010). In particular, for voting and conventional political activities, objective political knowledge is obviously a key causal component (e.g., Amadeo et al. 2002; Condon and Holleque 2013; Torney-Purta et al. 2001; Schulz et al. 2010), but in the prediction of unconventional or non-traditional political activities, internal political efficacy might be the more robust predictor (e.g., Kuhn 2006; Schulz et al. 2010). To complicate things, Johann (2012) and Westle (2011; 2012) provide evidence of the simultaneous predictive value of internal political efficacy, political interest, and political knowledge on conventional, as well as unconventional, political action using multivariate regression analyses (see also Gallego and Oberski 2012). The role of internal political efficacy versus political knowledge is thus equivocal especially because positive correlations between objective and subjective measures are of only a moderate extent (e.g., Maier 2000; Oberle 2012; Westle 2005; 2009; Westle and Johann 2010). This might support the assumption that both objective knowledge and internal efficacy have separate effects on political participation, maybe above and beyond mediated effects. What about political interest? Interest, motivation, and volition are strongly interrelated (e.g., Hidi and Renninger 2006; Ryan and Deci 2000). It is, therefore, not surprising that several authors point to the importance of political interest in behavioral politicization (e.g., Cohrs and Boehnke 2008; Hadjar and Becker 2006; Jugert et al. 2011; Schmid 2004; Schulz et al. 2010; Steinbrecher 2009; van Deth 2001).
Political Competences and Political Action: On the Mediation of Political Knowledge 6 Although political interest is a colored concept, it can be defined as the degree to which politics arouses a citizen s curiosity (van Deth 1990, p. 278). It comprises political awareness or attentiveness (cf. Zaller 1992) and includes a diffuse but general willingness to participate in political life without it being political behavior by definition. However, political interest is usually significantly correlated with internal political efficacy (e.g., Hadjar and Becker 2006; 2007; Reichert 2013; Vetter and Maier 2005). Thus, to avoid collinearity and because of the fact that internal political efficacy can be regarded as the subjective pendant of objective political knowledge, the focus was not on political interest but primarily on internal political efficacy. For the remainder, these variables will be referred to as cognitive politicization (cf. Reichert 2013). Variables Influencing Cognitive Politicization Political competences, including knowledge, usually increase with age, and it is those who are more highly educated as well as males 3 who are more cognitively politicized (e.g., Delli Carpini and Keeter 1996; Gaiser, Gille, and de Rijke 2010; Gidengil et al. 2004; Grönlund and Milner 2006; Krampen 1991; 2000b; Kuhn 2006; Maier 2000; van Deth 2013; Vetter 2006; Westle 2005; Westle 2012), even though Schulz et al. (2010) did not find an effect for gender on political knowledge. Furthermore, people have higher levels of political knowledge after political elections compared to before political elections (Maier 2009; Westle 2012). These variables may therefore be considered as controls in predicting political action and/or cognitive politicization. For respondents with Turkish migration history, Westle (2011; 2012) also identified a positive relationship between political knowledge and being born in the country of residence (i.e., Germany). Hypotheses To put it succinctly, the following assumptions are made: H1: Cognitive politicization variables correlate positively with at least legally accepted kinds of political action and action intentions in bivariate analyses. H2: When controlling for each other as well as socio-demographic variables, only a significant effect of political knowledge on voting remains; because of an overlap of both variables, political knowledge is mediated via internal political efficacy with respect to the other forms of legally accepted political action. 3 Adding to that the pattern for answers that are wrong, correct, or unknown, the answers vary depending on gender (e.g., Vetter and Maier 2005; Westle 2005; 2009; 2013). Furthermore, compared to men, women are less knowledgeable with regard to conventional political knowledge, but they gain better results than men in the policy dimension (Stolle and Gidengil 2010).
Political Competences and Political Action: On the Mediation of Political Knowledge 7 PRELIMINARY WORK: POLITICAL KNOWLEDGE AMONG UNIVERSITY STUDENTS In this chapter, the results of a panel study that was conducted in 2010 and 2011 are summarized, because those analyses constituted the basis for the analyses of the GLES data. Since details may already be found elsewhere (Reichert 2010; 2013), that preliminary study is only reported in brief. Sample and Method The first wave consisted of 76 university students from the local department of psychology who held German citizenship and who had acquired their A levels (Abitur) in Germany. The mean age of the respondents was 23 years (SD = 3.60), and most of the respondents were female (71%). Furthermore, nearly half of them had a migration background, whereas the families of 53% of them had lived in Germany for at least three generations. The mean income was around 525 Euros (SD = 269). Nine to ten months later, 41 participants were surveyed again to get information about their political behavior during that time. Between the students who had participated only in the first wave and the panel participants, there were no statistical differences in socio-demographics, but a smaller proportion of the panel sample had participated in conventional political action before the first time of measurement compared to students that were only surveyed the first time (10% vs. 29%) 4. insert figure 1 here Fig. 1 Confirmatory factor analysis for internal political efficacy (unstandardised factor loadings [λ i ], item discriminations [r it ], and fit indices). Items translated into English original item language is German. To measure internal political efficacy, a six-item index was used (see Figure 1). For each item, the students answered on a five-point scale (from 0 = not true at all to 4 = absolutely true; M = 2.17, SD = 0.91; Cronbachs α = 0.91). The political knowledge index consisted of fourteen items (see Figure 2), which were weighted by their item discrimination parameters and ranged from 0 to 4.476 (M = 4.07, SD = 1.68). Most of the items were single choice questions with three incorrect and one correct answer 5. insert figure 2 here Fig. 2 Item Characteristic Curves (ICC) for the 14 test items selected for analyses using a 2PL-Birnbaum model. 4 All significances and confidence intervals refer to p < 0.05, unless otherwise noted. 5 The test can be obtained from the author.
Political Competences and Political Action: On the Mediation of Political Knowledge 8 All respondents also stated whether they had voted in the German parliamentary election of 2009 (87% had); whether they had participated in conventional political activities (a dichotomized measure of three items, e.g., participated in an election campaign; 18% had); whether they had participated in unconventional behavior (a sum index with five items, e.g., participated in a citizens initiative; M = 1.82, SD = 1.31); and whether they had participated in non-normative political protest (dichotomized measure of six items, e.g., blocked a street for political reasons; 25% had). Moreover, every student answered how likely it would be for him or her to participate in one of those actions in the future (from 0 = definitely not to 4 = definitely). However, for electoral participation they were asked about any future political election in Germany (M = 3.72, SD = 0.81), and for conventional action only two items had been used (M = 0.91, SD = 0.82). For illegal activities, with regard to the social desirability of the responses, the students answered how much they sympathized with people taking part in those activities (from 0 = do not understand at all to 4 = completely understand; M = 1.69, SD = 0.76; unconventional activities: M = 2.19, SD = 0.98). In the second wave, the participants indicated whether they had voted in political elections between the first and the second measurement (54%; n = 13 because of missing opportunities to vote). Conventional political activity was measured using four items (10%) but unconventional (M = 1.54, SD = 1.31) and non-normative political action (18%) were measured with the same items as in the first survey. Important Results in Brief In the cross-sectional analyses, political knowledge only correlated significantly with the intention to participate in conventional political activities in the future (r =.24). A two-step analysis with political knowledge and internal political efficacy confirmed the assumption: When internal political efficacy was included in the analysis, political knowledge lost its significance (t(73) = -0.11, β = -0.01, p = 0.913) and only efficacy remained a significant predictor of readiness to participate conventionally (t(73) = 3.50, β = 0.45, p < 0.001; R² = 0.19). Mediation analysis using 5,000 bootstrap resamples and an alpha set at 0.05 confirmed a significant mediation (B = 0.13, SE = 0.04, CI [0.06 0.21]). Even under control for socio-demographic control variables (gender, age, migration background) and/or past conventional activity, only internal political efficacy remained significant. Consequently, internal political efficacy was the more proximal measure and covered, or mediated, the effect of political knowledge. A significant relationship between political knowledge and another kind of intended political action at time one, or a further mediation via internal political efficacy, could not be identified. Table 1 Bivariate correlations between political knowledge, internal political efficacy and political behavior insert table 1 here
Political Competences and Political Action: On the Mediation of Political Knowledge 9 In the longitudinal analyses, only the correlation between political knowledge (time one) and conventional political action appeared to be statistically significant (r = 0.41). In logistic regression analyses, neither political knowledge (OR = 2.17, Wald = 1.52, p = 0.218) nor internal political efficacy (OR = 13.11, Wald = 2.44, p = 0.118) was significantly related to conventional political action when controlled for each other. The adding of internal political efficacy was nevertheless statistically significant (χ 2 = 4.21, p = 0.040). The inclusion of other control variables (all of them with insignificant regression coefficients) did not change this pattern. Furthermore, mediation analyses indicated that political efficacy acted as mediator for knowledge (B = 0.27, SE = 0.12, CI [0.11 0.46]; consult e.g., Taylor, MacKinnon, and Tein (2008) for details about mediation analyses with insignificant direct effects). For the other action variables, no such mediation was detected. Bivariate correlations are depicted in Table 1, and Figure 3 summarizes the mediation analyses for conventional political action. insert figure 3 here Fig. 3 Internal political efficacy mediates the effect of political knowledge on (1) the intention to conventional political action (cross-sectional analysis; above) and (2) actual conventional political participation (longitudinal analysis; below) (unstandardized regression coefficients; : p <.001, : p <.01, : p <.05). Limits of the Preliminary Study Although political interest was measured using two items, it was not considered in the main analyses because of its strong correlation with subjective political competence (r = 0.83, p < 0.001). Because of the small sample size, there existed also a considerable danger of collinearity. It nevertheless should be noted that only political interest remained statistically significant if included in the cross-sectional mediation analysis. In the longitudinal analyses, however, none of the mentioned variables showed a significant regression coefficient. In summary, mixed results have been reported so far: at least political knowledge did not correlate significantly with all behavioral intentions, though with actual legally accepted political action (α < 0.10) in bivariate analyses (pro H1), but when controlling for internal political efficacy its significant effect on conventional action diminished (pro H2). POLITICAL KNOWLEDGE IN GERMANY: DIRECT AND MEDIATED EFFECTS ON POLITICAL ACTION
Political Competences and Political Action: On the Mediation of Political Knowledge 10 Procedure and Method The GLES long-term panel (Rattinger et al. 2012) merges data from pre- and post-election surveys conducted around the elections to the German Bundestag in 2002, 2005, and 2009. The population includes people with German citizenship aged 16 years and older who lived in Germany at the time of measurement. From the 3,263 people surveyed in 2002, 902 participated again in the second wave in 2005, and still 641 were surveyed in the third wave in 2009. For 436 persons, data were available for every wave. Because of the disproportionate sample design, the east/west weight was used in the following analyses; cases with mutations in age or gender were excluded from analyses. Cognitive politicization variables Two questions in the 2002 survey measured political knowledge explicitly: Which of the two votes in the federal election is of more importance? and, How many federal states does Germany have today? For the 2009 survey, only the first item was available. Furthermore, participants were asked to grade several nationwide important political parties on political issues like nuclear power, immigration of foreigners (only in 2002 and 2009), and European unification. They were given a seven-point scale, each of which was labeled only at the two extremes with opposing statements, for example, laws on immigration should be relaxed vs. laws on immigration should be made tougher. Since two items seemed to be inadequate in the measurement of political knowledge, the positioning of political parties was utilized in addition. For that reason, it was decided to take the respective election manifestos of each party as the basis for evaluating the correct party position. That seemed reasonable, because the GLES data were always conducted around federal elections, and the manifestos are written documents and thus are fixed information that voters can read to learn about the positions of the political parties. Then, for each wave and with reference to the election manifestos, two coders indicated which scale point most adequately represented the respective position of each party. Afterwards, both coders discussed the diverging coding most discrepancies were only of a difference of one scale point and came to an agreement in every contested case. Eventually, an additive index for political knowledge representing the percentage of correct answers was calculated ( don t know and non-responses were treated as wrong answers). Three efficacy items were compatible with those of the preliminary study and were able to be used as indicators for subjective political competence (from 0 = strongly disagree to 4 = strongly agree): Politics is such a complex issue that people like me cannot understand what is going on (reversed coding); I feel I could play an active role in a group dealing with political issues ; and, I am perfectly able to understand and assess important political questions. In 2005, however, only the first item was available; for 2002 (Cronbach s α = 0.70) and for 2009 (Cronbach s α = 0.65), mean indexes were calculated.
Political Competences and Political Action: On the Mediation of Political Knowledge 11 Behavioral politicization variables With respect to the four types of political action, several variables were available in the GLES dataset. Post-election participants were asked whether they had voted in the recent federal election (0 = no vs. 1 = yes), whereas pre-election participants indicated on a five-point scale how likely it was that they would vote in the forthcoming federal election (from 0 = I will certainly not vote to 4 = I will certainly vote). Moreover, independent of pre- or post-electoral measurement, retrospective information about voting decisions in the elections to the previous period of the German Bundestag were available (recoded to 0 = did not vote vs. 1 = voted). Conventional political action in the past was measured using two items: I visited election campaign rallies (five-point scale recoded to 0 = never and 1 = rarely to very often; not available in 2005), and membership in a political party (0 = not a member and 1 = member). The final index was coded 0 (no participation at all) and 1 (participation in at least one of the respective activities). The intention to participate in conventional political action could be measured via the item, Try to get support from a party/politician (from 0 = I would definitely not do to 4 = I would definitely do). The measurement of unconventional and non-normative past political action was difficult, since only one, single item comparable to the preliminary study was available: Membership in a citizen s initiative, For willingness to engage unconventionally, however, two items could be used: Contribute to a citizens initiative, and Participate in a legal demonstration (from 0 = I would definitely not do to 4 = I would definitely do). The same is true for willingness to act in a non-normative way: Participate in a violent demonstration, and Block the traffic (recoded because of skew distribution: 0 = I would definitely not do vs. 1 = I would likely not do to I would definitely do). Socio-demographic control variables Standard socio-demographic variables were used, like gender (0 = female vs. 1 = male; 51% male 6 ), age at time of measurement (M = 50.67, SD = 18.12), and migration history (0 = German citizen since birth vs. 1 = not German citizen since birth; 96% German citizens since birth). Furthermore, a household s total net monthly income was measured and recoded into eight nearly equally spaced intervals ranging from 0 = less than 500 Euros to 7 = at least 3 500 Euros (M = 3.26, SD = 2.04; i.e. a mean income of approximately 815 Euros). For the categories of class affiliation, two dummy variables were coded for working class (40%) and middle class (57%; remainders belonging to upper class, 2%) (0 = not affiliated to class vs. 1 = affiliated to class). The same is true for highest school graduation certificate, which was recoded into the dummies no or lowest formal qualification 6 The descriptive statistics given here apply to the first time of measurement (2002). Some frequencies do not sum up to 100% because of rounding.
Political Competences and Political Action: On the Mediation of Political Knowledge 12 (46%), and intermediary secondary qualification (29%; remainders with higher qualification, incl. university diploma, 25%) (0 = not true vs. 1 = true). The dummy variables for pre- vs. post-election date of interview (0 vs. 1; 49% preelection) as well as for region of residence (0 = Western Germany and 1 = Eastern Germany; 77% located in Western Germany) were also considered as controls in multiple analyses. Statistical analyses were conducted for the criteria of willingness to participate in political action, especially for cross-sectional relations. Actual participation after the time of measurement was also analyzed for voting behavior of people who were entitled to vote and conventional political action. To examine the bivariate relations (Table 2) in detail using multiple analyses (Tables 4 and 5), political knowledge and internal political efficacy were considered. The control variables mentioned above, including political activity before the respective time of measurement for cross-sectional analyses and behavioral intentions for longitudinal analyses, were also accounted for. With regard to actual voting and unconventional behavior, however, only age, gender and migration history, the intention to vote, and those socio-demographic variables which were significantly correlated with the respective criterion were considered in multiple analyses. The reasons for this are the high numbers of respondents who actually voted and who have performed unconventional political activities, respectively, on the one hand, and the aim nevertheless to make analyses compareable to the preliminary study. Since this article focuses on the cognitive politicization variables, controls are not always reported in the text. All multiple and mediation analyses were conducted in Mplus using the default setting (FIML) for dealing with missing data (cf. Muthén and Muthén 1998 2012) using 1,000 bootstrap resamples. Significant results of the multiple analyses correspond to bias-corrected confidence intervals, and mediation analyses were also conducted in cases without a significant direct link between predictor and criterion (cf. MacKinnon 2008; Taylor, MacKinnon, and Tein 2008). Correlations in Brief As depicted in Table 2, political knowledge did not always have a significant bivariate relationship to variables of behavioral politicization. There was not even a constant pattern for one of the relevant criteria. There was, however, a tendency for correlations between conventional action variables and knowledge to be of more importance than with other criteria. Internal political efficacy yielded significant correlations with most of the behavioral variables. As can be seen in Table 3, bivariate correlations between the variables of cognitive politicization were always highly significant and in most cases of moderate extent. Table 2 Means, standard deviations, and inter-correlations of the main variables
Political Competences and Political Action: On the Mediation of Political Knowledge 13 insert table 2 here Table 3 Bivariate correlations betwee political knowledge and internal political efficacy insert table 3 here Cross-sectional Regression Analyses: Willingness to Act as a Criterion Table 4 shows the regression coefficients of the cross-sectional analyses, as well as indirect effects on the respective criteria. Intention to political action was the main criterion, whereas internal political efficacy was considered as mediator of political knowledge and therefore regarded as intermediate criterion. Table 4 Results of the multiple regression analyses for willingness to participate politically (cross-sectional analyses) insert table 4 here Voting (pre-election sample) For the post-election sample, no comparable variable for intention to vote was available; therefore, for voting, the regression analyses could only be applied to the pre-election sample. As can be seen in Table 4, political knowledge had even a marginal significant direct negative effect on the intention to vote in 2005 when controlling for other variables. Internal political efficacy was always a significant and positive predictor of the intention to vote. However, only in 2002 and in 2005 it mediated the effect of political knowledge on voting intentions. The nonsignificant mediation in 2009 is due to the fact that political knowledge was no significant predictor of internal political efficacy in 2009 when control variables were included in the model. Furthermore, voting in the past had significant effects on the willingness to vote in the future. In 2002, also older people as well as women and people living in Western Germany were more likely to intend to vote in the future. Income had a positive effect on voting intention in 2009, wheres having achieved no or only the lowest eduational qualification had a marginally negative effect in 2005. Conventional action Political knowledge had positive effects on the intention to participate in conventional political action in 2002 and in 2005, though only marginally significant in 2002. In contrast, internal political efficacy was a significant positive predictor in 2002 and 2009, but not in 2005. Efficacy also mediated the effect of political knowledge in the 2002 analysis, but of course not with respect to the other two criteria.
Political Competences and Political Action: On the Mediation of Political Knowledge 14 Past political behavior stimulated the intention to act again in every analysis, whereas age yielded negative effects in all models. As already witnessed with regard to voting intentions, women and Western Germans were more likely to intend to engage in conventional political action in 2002. In addition, pre-election respondents as well as the higher educated were more likely to participate in 2002, as may be concluded by the negative dummies for education. People who assigned themselves to the working class or to the middle class were also more likely to participate in conventional political action according to the multiple analyses for 2002 (in 2009, the coefficient for middle class was marginally significant). Income had a positive effect in the model for 2005, and in 2009, native Germans were more likely to indicate a higher willingness to conventional political action. Unconventional action Regarding unconventional behavior, political knowledge was a significant positive predictor in 2002 and a significant negative predictor in 2009, even when controlling for other variables. Internal political efficacy always increased the intention to participate in unconventional political behavior. As in the analyses for voting intentions, efficacy was a significant mediator of political knowledge in 2002 and in 2005. Past political activity was a positive predictor in every wave, while age always yielded negative effects. Also, higher education increased the intention to engage in unconventinoal action in 2002, whereas respondents from the working class were more willing to participate in 2005 (in 2002 and in 2005, the coefficient of middle class is marginally significant and positive). People who were in the pre-election sample were also more prone to intend to participate in 2002, though this coefficient was only marginally significant. In 2005, people who lived in Eastern Germany indicated a higher intention to act in an unconventional way, and income had a positive effect in 2005as well. Non-normative action At first sight, the pattern for the willingness to act in a non-normative way is similar to that for conventional political action: In 2002, political knowledge which itself had a significant positive effect was mediated via the positive predictor internal political efficacy. In the subsequent wave, a marginally positive effect of political knowledge emerged, whereas internal efficacy was no significant predictor in 2005. In the last wave, knowledge did not have a significant effect, whereas efficacy was a positive predictor of the intention to engage in non-normative political behavior in when controlling for other variables. Again, age always yielded negative effects. Higher education increased the intention in 2002 and in 2005, and respondents from the working class as well as people from the middle class were more likely to intent to act illegally in 2002 (marginally significant coefficient for working class). Income had a negative effect in 2002, when people from
Political Competences and Political Action: On the Mediation of Political Knowledge 15 Eastern Germany rather tended to act in a non-normative way. Men indicated a higher willingness to use non-normative political behavior in 2005. Summary of cross-sectional analyses Bivariate analyses indicated that internal political efficacy was always a positive correlate of behavioral intentions, and in many cases political knowledge was, too (pro H1). In multiple analyses, however, political knowledge still remained a (marginally) significant predictor (α < 0.10) even when controlling for internal political efficacy and further variables. Those effects were not always positive, but appeared more often not on voting but on other criteria (contra H2). Internal political efficacy was nevertheless a positive mediator of political knowlegde in 2002, and with regard to voting and unconventional political behavior even in 2005 (pro H2). Unfortunately, political knowledge did not have a significant effect on political efficacy in multiple analyses in 2009, so that efficacy was not a mediator then. Hence, we found mixed results for H2. Longitudinal Analyses: Explaining Actual Political Behavior Table 5 summarizes the results of mediated regression analyses for the panel samples with regard to voting, conventional behavior and unconventional political participation. These results are described in the following sections. Table 5 Results of the multiple regression analyses (longitudinal analyses) insert table 5 here Voting Results of longitudinal analyses using the pre-election sample can be inspected in Table 5. For the post-election sample 7, no significant bivariate correlation between political knowledge and voting was found, and none of the mediation analyses yielded any allusion to a significant effect. In the pre-election sample, internal political efficacy never had a significant effect on voting, whereas in the 2002 analysis political knowledge kept a significant and positive effect even under control of socio-demographic variables. The intention to vote significantly increased the likelihood to vote in 2002 and in 2005 when controlling for predictors measured in 2002, whereas in the analyses for 2005 using predictors measured in the same year, age had a marginally significantly positive effect on voting, and being a working class member had a negative effect. 7 Here, no behavorial control variable could be used.
Political Competences and Political Action: On the Mediation of Political Knowledge 16 As political efficacy had a positive effect on voting intentions, but not on voting itself, a sequential mediation model was conducted subsequently. For the purpose of testing whether efficacy had an indirect effect on voting being mediated by the intention to vote, the largest sample of 2002 was used. The respective model is depicted in Figure 4 and includes all control variables (exclusive of dummy variables for class affiliation). insert figure 4 here Fig. 4 Sequential mediation of political knowledge by internal political efficacy and intention to vote on voting in the 2002 federal elections (predictors measured in 2002). Note: Unstandardized regression weights; for mediated regressions, direct effects are depicted in parentheses. Significant coefficients are indicated as follows (with respect to bias-corrected confidence intervals): p < 0.001, p < 0.01, p < 0.05, p < 0.10 (ns = not significant). For more information see text. Not surprisingly, political knowledge had a significantly positive effect on internal political efficacy, and political efficacy siginificantly increased the intention to vote. Furthermore, the latter predicted actual voting behavior in the pre-election sample (positive effect). Political knowledge had no direct effect on the intention to vote, but a positive effect on voting. Also, political efficacy had no direct effect on voting, but an indirect positive effect via the intention to vote. Consequently, a sequentially mediated effect of political knowledge on voting via the mediators internal political efficacy and intention to vote was affirmed. With respect to control variables, immigrants were more likely to vote (B = 2.63, CI = [0.20 3.53]). While age increased the intention to vote (B = 0.01, CI = [0.01 0.01]), people who lived in Western Germany were more willing to cast their ballot (B = -0.23, CI = [-0.37-0.11]) and also more politically efficacious (B = -0.18, CI = [-0.29-0.07]). Those who had no or only the lowest educational qualification were less likely to intend to vote (B = -0.15, CI = [-0.27-0.02]) and felt less politically efficacious (B = -0.18, CI = [-0.80-0.58]). Internal political efficacy was also higher among men (B = 0.41, CI = [0.32 0.50]) and for people who did not have an intermediary secondary qualification (B = - 0.17, CI = [-0.28-0.05]), hence, higher formal education increased internal political efficacy. Conventional action In the prediction of participation in conventional political action, political knowledge never was a significant predictor in multiple analyses, but internal political efficacy and the intention to engage in conventional action always were significant an positive predictors (cf. Table 5). Internal political efficacy appeared to be a mediating variable, i.e. political knowledge always had indirect effects on conventional behavior. Among all control variables, older people
Political Competences and Political Action: On the Mediation of Political Knowledge 17 were significanlty more likely to engage in both analyses using predictors measured in 2002. Between 2005 and 2009, participants in the post-election sample were more likely to act politically in a conventional way (both models). There was also a marginally significant effect of region in the analysis for participation in 2005. As it was shown that the effect of internal political efficacy on actual behavior might at least in part even be mediated by action intentions, such an additional model was also conducted. Again, predictor variables measured in 2002 were used for subsequent participation between 2002 and 2005. However, this time it was possible also to control for past conventional behavior, so that results would predict changes in actual participation. This is possible as the intention to participate in conventional action was significantly predicted by efficacy controlling for previous behavior in the 2002 model (cf. Table 4). The results are shown in Figure 5. insert Figure 5 here Fig. 5 Sequential mediation of political knowledge by internal political efficacy and intention to participate in conventional political action on actual conventional behavior between 2002 and 2005 (predictors measured in 2002). Note: Unstandardized regression weights; for mediated regressions, direct effects are depicted in round parentheses; the indirect effect of political knowledge by intention to participate on actual behavior is given in squared parentheses. Significant coefficients are indicated as follows (with respect to bias-corrected confidence intervals): p < 0.001, p < 0.01, p < 0.05, p < 0.10 (ns = not significant). For more information see text. The analyses yielded a significant direct and positive effect of political knowledge on internal political efficacy, a marginally significant positive effect on the intention to participate conventionally, but no direct effect on actual behavior. Political efficacy had a positive direct effect on the intention to engage in conventional political action, but not on actual participation. Note that this deviation from the analyses presented above (cf. Table 5) is due to the inclusion of past political behavior in the model presented in Figure 5 and, thus, not too surprising. Finally, the action intention predicts actual conventional behavior. The results of mediation analyses show that the effect of internal political efficacy on conventional action is mediated by the respective action intention. For political knowledge, which had a marginally significant direct effect on the intention to participate, an additional significant mediation by political efficacy appeared. Despite these two significant mediated effects, however, the sequential mediation of political knowledge via efficacy and action intention
Political Competences and Political Action: On the Mediation of Political Knowledge 18 on conventional participation is only of marginal, but positive significance. In addition, there existed a marginally significant simple mediation of political knowledge by the action intention on actual behavior. Among control variables, past conventional political participation predicted actual participation (B = 1.64, CI = [1.23 1.95]), the intention to participate in conventional action (B = 0.37, CI = [0.24 0.49]) and internal political efficacy (B = 0.36, CI = [0.29 0.44]). Age increased the likelihood to actually vote (B = 0.02, CI = [0.01 0.03]), but it also led to a lower intention to engage in conventional political action (B = -0.01, CI = [-0.02-0.01]). The intention to participate was higher in the pre-election sample (B = -0.11, CI = [-0.21-0.01]) and among people from the Western part of Germany (B = -0.24, CI = [-0.37-0.12]). Moreover, women were more willing to vote in the future (B = -0.12, CI = [- 0.21-0.01]), but reported also lower internal political efficacy (B = 0.38, CI = [0.32 0.44]). Higher education was positive for the intention to vote (dummy for no or lowest educational qualification: B = -0.29, CI = [-0.45-0.14]; dummy for intermediary secondary education: B = -0.17, CI = [-0.31-0.03]) as well as for internal political efficacy (dummy for no or lowest educational qualification: B = -0.49, CI = [-0.58-0.40]; dummy for intermediary secondary education: B = -0.18, CI = [-0.27-0.10]). On the contrary, lower class affiliation increased the intention to engage in conventional action (dummy for working class: B = 0.53, CI = [0.14 0.91]; dummy for middle class: B = 0.52, CI = [0.16 0.89]), however, it decreased internal political efficacy (dummy for working class: B = -0.57, CI = [-0.77-0.37]; dummy for middle class: B = -0.34, CI = [-0.53-0.15]). Eventually, political efficacy was higher in the post-election sample (B = 0.07, CI = [0.01 0.13]). Unconventional action Unconventional political action in 2005 is predicted by internal political efficacy, but not by political knowledge of which the effect was in fact mediated by efficacy. The latter was not a significant variable in the model for unconventional action in 2009, in which political knowledge had a marginally significant and positive effect on the criterion. In the 2009 analysis using predictors measured in 2005, neither knowledge nor efficacy predicts unconventional behavior. Age always increased the likelihood to participate, but neither of the remaining control variables was a significant predictor (except a marginal signficant coefficient of intermediary secondary qualification in one model). Another sequential mediation model was calculated for unconventional participation, again with predictors measured in 2002 and the 2005 criterion, because only in this longitudinal regression analysis, political efficacy had a significant effect on respective political action. Only control variables mentioned in Table 5 were used for the dependent variable unconventional action, but for criteria measured in 2002, all controls given in Table 4 could be used.